The AI hyperscalers are lumped together for obvious reasons, but after the four largest reported their earnings on Wednesday night, it became abundantly clear that one of these big-spending giants is not like the others.
Shares of Meta Platforms Inc., the night’s biggest loser, were down 7% in after-hours trading as Mark Zuckerberg counted the cost of declaring yet another capex increase. The latest estimate calls for as much as $145 billion for the year, up from the last forecast of up to $135 billion.
That spending is in the same arena as the other cloud businesses. Except, Meta isn’t a cloud business. So when it breaks the news that it’s going to need to spend even more, it can’t point to healthy growth in cloud sales from AI — because obviously there aren’t any. Investors are right to question what Meta will get in return for all that AI investment when ultimately its only client is itself.
Contrast Meta’s evening to one of the night’s winners, Alphabet Inc.’s Google. As well as demonstrating that its legacy business is fending off the threat from AI, the company’s 63% cloud revenue growth made its projected capex increase, to as much as $190 billion for the year, far easier to swallow. Like Meta, Google has its own AI chips, except they are being sold to other companies as well as being used to run Google’s own AI workloads more efficiently. Its shares gained 4% in after-hours trading.
Also consider Amazon.com Inc., which reported higher spending as well but did so off the back of better-than-expected revenue growth for its cloud unit — 28%, the fastest since 2022. That’s before it brings on its new partner, OpenAI, alongside existing client Anthropic. Both AI labs are the front-runners in offering the coding functionality that has been AI’s most significant break-out use case so far.
Even Microsoft Corp., which was been held back by worries about OpenAI’s future, managed to use its earnings call to restore some confidence with positive signals on cloud growth and “seats” for its AI-enhanced tools.
Yes, all three companies still face concerns about the ultimate return on investment from all this spending, but at least when you’re a cloud company you can draw a clearer line between spending out and revenue in.
But to reiterate, Meta isn’t a cloud-computing company, so it can’t do that. That’s stating the obvious, of course, but the point is maybe it shouldn’t be spending like one. Apple Inc. has enlisted Google’s help to make Siri smarter and plans to offer an array of AI models on its iPhone, much in the way users can tap into different apps or search engines. Apple’s capex, incidentally, is roughly 10 times smaller than Meta’s, thanks to not feeling the need to build a data center comparable to the size of Manhattan.
Meta could have opted to work with partners. As the world’s largest social network — 3.56 billion daily active users — Meta would have had AI labs scrambling to win the contract to power the AI on Facebook, Instagram and WhatsApp. But Zuckerberg’s desire to “own the tech stack” and handle everything is at the heart of his competitive spirit and Meta’s ballooning spending.
It’s too early to say he’s misguided in his AI aims. There’s a reason he’s worth $237 billion. There are, of course, cost benefits to be had from owning the model you offer to billions of users. But at times, Wednesday’s wishy-washy earnings call had more than a whiff of “next year will be the year of the metaverse” about it. The utility of Zuckerberg’s “personal” AI still seems only loosely defined, and his social network isn’t really useful for much at all.
When asked about what signs he’s looking for over the next 12 to 24 months that his AI work is on track, Zuckerberg offered: “Basically, we invest in advance to build leading models, then we convert that into leading products.”
He continued: “I don't think we have a very precise plan for exactly how each product is going to scale month over month, or anything like that. But I think we have a sense of the shape of where these things need to be.”
He then cited the “quality” and “trajectory” of other companies’ models as reasons to feel confident in the potential of his own. It wouldn’t be the first time he’s had to catch up to those out-innovating him. Snap Inc.’s Chief Executive Officer Evan Spiegel last year cheekily had his title on LinkedIn set to “VP of Product @ Meta” on account of the company’s rampant copying of Snapchat features.
Zuckerberg’s giant Xerox is surely meeting its limit with AI. It’s a very tall order for Zuckerberg to pull a Steve Jobs and be late to the market but with a product so superior it ends up not mattering, which is the reality of the goal that has been set. At the very least, if Zuckerberg’s going to keep expecting investors to be on board with a hyperscaler budget, he needs to come up with something that resembles a hyperscaler business model.
A message from Advisor Perspectives and VettaFi: Discover something new! Click here to register for our upcoming webcasts.
Bloomberg News provided this article. For more articles like this please visit
bloomberg.com.
Read more articles by Dave Lee
